Day 2 – Opening AI Panel – 8:35-9:30 AM
Room: Affectiv
Stop Chasing Tools. Start Building Output.
AI Panel: Jeremy Kohler, Austin Winfield, Ryan Kohler & Ryan Harris · RizeCon 2026 · Pocatello
Half the room was recording calls. Half of those were doing nothing with the recordings. The panel stopped to ask — and the number that came back was exactly what three of the sharpest AI practitioners in the room expected. The conversation that followed wasn’t a pitch for a new platform or a tutorial on which model is best. It was a direct, occasionally blunt breakdown of why most businesses are using AI wrong, what the difference between automation and AI actually is, and why the quality of what you put in determines everything about what comes out.
Moderated by Jason Kohler, the panel brought together Austin Winfield (co-founder, Antogram), Ryan Kohler, and Jeremy to cover the ground between “where do I even start” and “how do I orchestrate this at scale.” The answer to both questions, it turned out, is roughly the same: treat it like a new hire, give it context, define what done looks like, and build from there.
What they covered
Start with one repeatable problem. The unanimous answer to where beginners should start: pick one thing you do every single week, something low-stakes enough that if the output is wrong it won’t hurt the business, and automate it incrementally. Not the whole workflow — just one step. Draft the email, review it, send it manually. Once that’s working, get it to send. The mistake most people make is trying to boil the ocean: automating everything at once, solving every problem in the business simultaneously, and then blaming the tool when nothing works. Jeremy’s version of this principle was direct — go figure out one problem that is measurable, incrementally automate it, have metrics, and adjust based on results.
AI is not automation — know the difference. Ryan drew the line clearly: automation is what happens after you’ve achieved near-perfection in a workflow. It means you can put something in and predict exactly what comes out. If you automate before you can predict the output, you’ve handed a teenager the company credit card and told them to send emails on your behalf. The hunger to scale before the process is nailed — what the book Nail It Then Scale It names precisely — is one of the most common failure modes in AI adoption. Scale without awesome is just a mess.
Your input quality is everything. The models are not the variable. Ryan made the challenge explicit: give him the worst language model available, and he’ll outperform anyone in the room using a better one, because he indexes on the quality of his side of the equation. The simplest mental model for what a language model does: it finishes your sentence. It has no memory of you, no context about your business, no idea what your voice sounds like — until you give it that. A vague seven-word prompt is like introducing yourself to a brilliant stranger who just wiped their memory and expecting them to sound like you.
Context files and prompt frameworks. Austin runs Antogram, which builds AI-powered shopping experiences for brands — and the first thing they do with every client is centralize all product data. The lesson from that work: not all information is helpful. Feed it what’s relevant and timely. Don’t drop in sensitive or confidential data. Don’t dump everything you have. Think about what a new employee would need to produce the output you want — example posts, past work, success criteria — and feed it exactly that. Jeremy’s framing: think about what you’d give a social media manager on day one. Claude can’t be your social media manager if you haven’t told Claude what your past manager was producing.
The voice of your customer is the most valuable data. Ryan pushed back on the conventional definition of data — most people think of numbers. The most valuable data in any business is the voice of the customer stating their pain, in their own words. That’s not in a spreadsheet. It’s in recorded sales calls, in support tickets, in meeting transcripts. Austin’s company queries their recorded sales demos to answer product prioritization questions: how many prospects have asked for this feature? That replaces the loudest voice in the room with the actual voice of the market.
Talk to it — don’t type. Ryan’s single strongest recommendation for getting better output: use voice. Humans have been storytellers for millennia. Typing reduces effort. Speaking in context gives the model your language, your backstory, your reasoning — everything it needs to sound like you. Open Voice Memo, record your context, let it finish the sentence. The most powerful prompt in the room: “Here’s the project. Here’s where I want us to get. Now — what do you need to know about me and my business? Interview me. You take the lead.”
AI as instrument, not hammer. Ryan made the distinction between a tool and an instrument. A hammer gets used the same way by everyone. A guitar gets played differently by every person who picks it up — and the more you play, the more your art develops. Watching how other people use it, learning their framework, and then making it your own is exactly how mastery works. His practice method: use it on things that don’t matter. He spent an evening building a music app for a friend’s kids, writing an album from their stories, burning tokens on something low-stakes — and the underlying skill of prompting for emotional resonance is the same skill that makes a great blog post.
The leadership gap. Ryan dropped a line that landed: leaders at businesses in the $500K–$20M range have almost certainly used generative AI less than their lowest-level employee. If that’s true, it’s a leadership crisis. Employees will ask how their future employer uses AI — and the answer will matter to their decision. You can’t implement what you don’t know.
What attendees got
Ryan closed with a practical habit for teams: get around a table with devices and ask each person, “What’s the craziest, coolest thing you did with AI this week? Show us. What was the prompt?” That’s the original definition of a mastermind — not a guru, just five people sharing context, and an extra mind showing up. The suggestion for RizeCon attendees specifically: have those conversations today, in the hallways and between sessions, with vendors and fellow attendees. The learning compounds fast when you’re watching other people play the instrument.
One exchange that landed
An attendee asked what the biggest dead ends were for people just starting out. Ryan didn’t hesitate: chasing tools and chasing savings. The YouTuber telling you to switch to the newest model has one goal — views. He’s never run a company with employees or a payroll. His content is designed to get clicks, not to grow your business. And trying to vibe-code your own CRM to save $300 a month when that same tool could generate $30 million in revenue is exactly the kind of thinking that keeps businesses small. Chase revenue, not savings. The math never works the other direction.
“Scale without awesome is just a mess.” — Ryan Kohler
“Standardize the predictable so you can humanize the exceptional.” — Jeremy Kohler
“AI is not the solution. AI is a lever that can apply to a solution.” — Austin Winfield
About the panelists
Ryan Harris moderated the session and brings fifteen years running a digital agency to his work helping businesses implement AI strategically. Austin Winfield co-founded Antogram, which builds AI-powered customer-facing shopping and recommendation experiences for brands in ski, technical computing, and adjacent industries. Ryan Kohler is a serial entrepreneur who has scaled companies and now operates as one of the more advanced AI orchestration practitioners in the room, running multi-model workflows through Manus. Jeremy Kohler led software product teams for over a decade before bringing that same incremental build philosophy to AI implementation.